Abstract: It is hard to extract the target information of the license plate image in complex environments such as low light illumination and strong background noise. To solve this problem, a target information detection method for noisy low-light image is proposed based on cascaded Duffing oscillator and adaptive maximum entropy threshold segmentation. The license plate image is enhanced by the cascaded Duffing oscillator. In order to effectively detect the target information, an adaptive maximum entropy threshold segmentation method based on genetic algorithm is used to process the license plate image enhanced by cascaded Duffing oscillator. Compared with the improved histogram equalization approach and the image enhancement method based on Retinex, the image processing method based on the Duffing oscillator has obvious advantages in image enhancement and image noise reduction. Through experimental verification, compared with the original image, the method proposed in this paper improves the contrast by 58.0581 and the peak signal-to-noise ratio by 17.141 8 dB, which demonstrates that the proposed method can detect the target information of noisy low-light the license plate image effectively.

Key words: license plate image enhancement, low light, Duffing oscillator, maximum entropy threshold segmentation

摘要： 针对夜间照度低、背景噪声强等复杂环境下车牌信息提取难的问题,文中提出一种基于级联Duffing振子系统和自适应最大熵阈值分割的低光照含噪图像目标信息检测方法。通过建立级联Duffing振子系统,实现对车牌图像的增强,有效地检测出车牌信息,并利用基于遗传算法的自适应最大熵阈值分割方法对增强后的车牌图像进行后处理。通过实验验证,相比原始图像,文中提出的方法在对比度上提升了58.0581、在峰值信噪比上提升了17.1418dB,可有效检测出低光照含噪车牌信息。

关键词: 车牌增强, 低光照, Duffing振子, 最大熵阈值分割